OpenE2K
/
gcc
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

re PR target/7042 (cris-elf miscompiles line_hints in ghostscript - "return" delay slot bug) 

PR target/7042
	* config/cris/cris.c (cris_simple_epilogue): Kludge around reorg.c
	bug by saing that the epilogue isn't simple if there's a
	non-empty current_function_epilogue_delay_list.
	* config/cris/cris.md ("return"): Add sanity check asserting that
	current_function_epilogue_delay_list is empty.

From-SVN: r54721
This commit is contained in:
Hans-Peter Nilsson 2002-06-17 22:06:34 +00:00 committed by Hans-Peter Nilsson
parent e12a22a625
commit fd4858461e
3 changed files with 22 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
gcc/ChangeLog
View File

@ -1,3 +1,12 @@
2002-06-17 Hans-Peter Nilsson <hp@axis.com>
PR target/7042
* config/cris/cris.c (cris_simple_epilogue): Kludge around reorg.c
bug by saing that the epilogue isn't simple if there's a
non-empty current_function_epilogue_delay_list.
* config/cris/cris.md ("return"): Add sanity check asserting that
current_function_epilogue_delay_list is empty.
2002-06-17 Richard Henderson <rth@redhat.com>
* config/i386/i386.h (BIGGEST_FIELD_ALIGNMENT): Define instead

8
gcc/config/cris/cris.c
View File

@ -2042,6 +2042,14 @@ cris_simple_epilogue ()
|| current_function_outgoing_args_size
|| current_function_calls_eh_return
/* Kludge for 3.1: when reorg changes branches to the return label
into return insns, it does not handle the case where there's a
delay list for the epilogue: it just drops the insns in
current_function_epilogue_delay_list on the floor, resulting in
invalid code. We kludge around it in that case by saying that
we don't have a simple enough epilogue to use return insns. */
|| current_function_epilogue_delay_list != NULL
/* If we're not supposed to emit prologue and epilogue, we must
not emit return-type instructions. */
|| !TARGET_PROLOGUE_EPILOGUE)

5
gcc/config/cris/cris.md
View File

@ -3394,6 +3394,11 @@
/* Just needs to hold a 'movem [sp+],rN'. */
char rd[sizeof (\"movem [$sp+],$r99\")];
/* Try to avoid reorg.c surprises; avoid emitting invalid code, prefer
crashing. This test would have avoided invalid code for target/7042. */
if (current_function_epilogue_delay_list != NULL)
abort ();
*rd = 0;
/* Start from the last call-saved register. We know that we have a